This proposal describes two separate but closely related research projects using the newly released Health and Retirement Survey Project 1, conducted by the PI John Rust, will use the HRS data to estimate, test, and simulate a detailed dynamic programming (DP) model of retirement behavior that accounts for the sequential nature of the retirement process and individuals' subjective uncertainty about mortality, family status, health status, health expenditures, and income from employment, pensions, and assets. The model treats individuals as rational expected utility maximizers who follow an optimal retirement strategy that accounts for uncertain shocks such as the risk of falling into ill-health or incurring catastrophic uninsured health care expenditures. A previous version of this model has been estimated and tested using the Retirement History Survey (RHS) and is attached in appendix 1. Project 2, PhD dissertation research by the project assistant Maria Perozek, will use the HRS data to determine the impact of uninsured risks on the saving and transfer behavior of the elderly. In particular, this project will focus on the effect of informal family risk sharing arrangements on the saving of the elderly. Project 1 will make extensive use of the HRS to construct new variables needed to develop an expanded and more realistic version of the DP model. Once the estimated model is fully tested and validated, it will be used to predict the impacts of various policy changes on retirement behavior including the effects of national health care reform on early retirements, the effects of Americans with Disability Act and age discrimination legislation on the work incentives of older and disabled individuals, the 1983 Social Security reforms, and tightening of benefits and eligibility criteria for disability insurance. The specific enhancements to the DP model made possible by the HRS data include construction of more detailed measures of health status, much richer information on job demands and hours of work restrictions, the ability to distinguish between voluntary vs. involuntary job separations, the ability to model individuals' decisions to apply for unemployment insurance and disability benefits, improved data on health insurance coverage, and family transfers. Finally the HRS has detailed data on private pension plan provisions. This will allow us to include a significant fraction of individuals with private pensions who were excluded from the previous DP model due to lack of information on private pension provisions in the RHS. Project 2 will formulate a dynamic life cycle model which incorporates mortality risk, uncertain medical expenses and the possibility of informal family insurance. This project will exploit the comparative richness of the HRS data on the subjective risk of mortality and nursing home use, the health conditions and general health status of the respondent, intergenerational transfers of money and time, and formal health and life insurance coverage of the elderly. These data will be used to estimate and test the model, and to evaluate the effects of policies aimed at eliminating the uninsured risks facing the elderly population.